Frontier Lab Researcher: We Still Don't Know How to Reliably Instil Values in AI
davidmanheim · x · 2026-11-15
Neel Nanda, an alignment researcher at a frontier lab, responded plainly: "I work at a frontier lab, we do not know how to reliably instil values into the AIs, this is the scientific consensus."
A rare on-the-record admission from someone inside a frontier lab, useful as direct evidence of where value alignment actually stands.
More from AGI Musings
- 'Wordcel' Celebrates Using Plain-Language Poetry to Command Computers — StewartalsopIII · 2026-11-15
- "Most People Caring About AI Safety Really Mean Caring About Being in Control" — nptacek · 2026-11-15
- Manheim: narrow exceptions don't negate deontological constraints — davidmanheim · 2026-11-15
- Manheim: saying consequences justify breaking NDAs is an appeal-to-consequences fallacy — davidmanheim · 2026-11-15
- Open models aren't inherently safer—just 4-7 months behind closed ones, argues new post — BlackHC · 2026-11-15
- Manheim: honoring commitments is deontological — and consequentialists benefit too — davidmanheim · 2026-11-15